[sdiy] Variable rate waveform playback in NED synclavier

[rburnett at richieburnett.co.uk]

Hi Tom and all,

> On the left hand side of the
> diagram are the two Sample Rate Generators, which seem to take a 
> clock
> signal Fclk and multiply it by a fraction N/M, where N and M are 
> 8-bit
> numbers. How is this done?

I'd guess it's done with a PLL that has a divider in the feedback path 
(divide by N), followed by another divider outside the loop (divide by 
M.)  The fact that the first divider is inside a negative feedback loop 
would make its effect multiplicative on the VCO's output frequency.  The 
order of the steps might be swapped around, so that the crystal 
frequency is first divided by M, then multiplied by N inside the PLL to 
keep the intermediate frequency from becomming excessively high!

> This seems a bit dirty for NED.

Why is it dirty?  It's a very common technique used to give clean 
output signals from PLLs for all sorts of things like radio 
transmitters, receivers etc.  It may be that the 16-bits worth of 
"sample rate generator data" are only changed infrequently like when 
changing octave or changing patch so a momentary un-locking of the PLL 
for a millisecond or two would likely not be noticed.  After all, the 
N/M multiplier with integer N & M is quite a course method of frequency 
control similar to the 16'/8'/4' octave selection on juno DCOs.  Any 
fine modulation and pitch-bend would surely be done via the phase 
increment to the NCO?

> If not, how? I can't see how this bit works.

Another possible method would be harmonic multiplication of the 
original crystal frequency.  Since the crystal frequency is fixed and 
well known, it would be very easy to generate integer multiples of this 
frequency using harmonic doublers, triplers etc.  Again, this is a 
common technique in RF engineering for frequency multiplication.  It 
only requires a biased transistor and an LC tuned circuit for each 
frequency multiplier stage, and works very well when all of the 
frequencies are fixed.

> Secondly, once this new sample rate is generated from the Fclk, it's
> fed to a Phase Angle Incrementer, which I take it is what we'd call 
> an
> NCO. This seems to have a 8-bit frequency increment. So the final
> frequency is some weird combination of the variable sample rate
> Fclk*N/M and the freq inc - does this seem right or am I miles off
> here?

It sounds right to me.  Save for the weird N/M multipier the rest is 
just like a dual-channel DDS (direct digital synthesis) chip with its 
NCO, waveform lookup ROM and DAC.

> This is just a case of me being curious about famous instruments from
> the past. These things are still revered by many, so I'd like to know
> how they did what they did.

It's all good stuff, and you never know when an old trick from the past 
will help you solve a new challange with modern parts, or eeek out that 
extra bit of performance.  Will be interested to hear others replies 
about this.

-Richie,